Generative AI and the Magical Fable

After months of impatience and doubts within part of the ecosystem— “the hype is here, the revenue is not” had even become the catchphrase of financial analysts during the earnings presentations of listed American companies— Big Tech has finally managed to generate revenue through Artificial Intelligence.

Last quarter, Microsoft Cloud exceeded $33 billion in revenue. As for Google Cloud’s revenues, they increased by 22% compared to last year. In other words, companies are paying for generative AI services from cloud providers.

However, these results are not an end in themselves. Indeed, while we can applaud the nascent economic impact of generative AI for the makers of large language models, we must acknowledge that most companies still have a long way to go. According to QuantumBlack, less than 10% of companies have seen the effect of generative AI on their EBIT! This concerning figure echoes the presentation of the report by the Interministerial Commission on AI a few weeks ago.

On this occasion, its chairman, Philippe Aghion, quantified the increase in GDP by AI (“a range from 250 billion to 420 billion euros by 2034”) but refrained from doing the same with generative AI because “its recent popularity makes it difficult to have sufficient perspective.” Obviously, none of its members doubt the potential of this disruptive technology. However, this caution cannot be separated from the slow value creation by companies through AI.

Generative AI is not yet “more profound than fire, electricity, or anything that we have done in the past.”

In mid-February, the specialist media outlet The Information published an article titled “Amazon and Google Temper Expectations Around Generative AI.” The authors relayed testimonies from executives wondering when they would see the benefits of these “software programs supposed to automate repetitive tasks.” Contrary to the rhetoric that accompanied its invention, generative AI is not magical. It is not yet “more profound than fire, electricity, or anything that we have done in the past” as some Silicon Valley executives claim. For now, it is a technology that needs to be subjected to a methodical approach to create value. This involves identifying a limited number of high-value-added use cases. That is, the singularities that underpin a company’s competitiveness.

In this regard, the press is one of the most mature industries on this topic. Many newsrooms have already identified specific use cases that allow them to differentiate and improve their competitiveness without compromising their editorial staff. For example, the Financial Times is currently testing the “Ask FT” service, which allows its readers to access all the media’s articles from several decades via natural language queries. In this case, the specific use case for the British newspaper is the valorization of an archive database spanning several decades. Commenting on this innovation, Lindsey Jayne, the project manager, stated that it was not about “jumping on the hype train because, otherwise, the public would treat it as a gadget they would eventually tire of.”

Adopting a methodical approach will not necessarily lead to the use of generative AI. In some cases, “traditional” AI is more than sufficient. As was widely discussed at the last Davos meeting, many leaders still struggle to fully exploit the potential of AI. For them, generative AI could turn out to be a financial sinkhole and an operational disappointment.

We must not lose sight of Artificial Intelligence’s purpose—whether generative or not—which is to transform. However, this transformative capacity is not a given. To think otherwise is a mistake. To claim the contrary is a lie.

‍